Late Endosomes Act as mRNA Translation Platforms and Sustain Mitochondria in Axons

1Local translation of mRNAs is a highly conserved mechanism that allows spatial and temporal control of the proteome at a subcellular level. This regulation is particularly important in neurons, where tight control of protein localization allows highly compartmentalized functions far from the cell soma. mRNAs made in the nucleus associate directly with RNA-binding proteins (RBPs) and are transported as ribonucleoprotein particles (RNPs) to distal subcellular locations for local translation. However, the mechanisms regulating precise RNP localization and nascent protein synthesis at defined neuronal sites remain largely unknown.

2The endosomal pathway has been implicated in mRNA localization and translation in the fungus Ustilago maydis, where the polarized growth of long hyphae requires endosomal-dependent transport of the RBP, Rrm4, and translation of its cargo septin mRNA. In neurons, membrane trafficking relies on endosomes, which carry a range of proteins and lipids for targeted delivery. The endosomal pathway internalizes cargos from the cell surface, regulates their storage and their recycling, or sends them to lysosomes for degradation. In addition to their role in trafficking, endosomes operate as platforms where diverse intracellular signaling cascades can be activated or sustained. Two of the main players of this endosomal system are the early and late endosomes that can be distinguished by their associated Rab guanosine triphosphatases (GTPases); Rab5 coordinates clathrin-dependent endocytosis and biogenesis of early endosomes and their fusion, whereas Rab7 regulates the transport and maturation of acidic late endosomes as well as their fusion with lysosomes.

3Here we show that RNPs associate with motile Rab7a endosomes along retinal ganglion cell (RGC) axons. RNP-bearing Rab7a endosomes frequently dock at mitochondria, where they serve as hotspots for de novo protein synthesis. Disruption of Rab7a function by expression of Charcot-Marie-Tooth disease type 2B (CMT2B)-linked Rab7a mutants leads to impaired local protein synthesis, mitochondrial dysfunction, and loss of axon integrity.

4RBPs and ribosomes associate with motile endosomes in fungal hyphae, raising the possibility that endosomes are involved in RNA granule trafficking in other cell types with elongated profiles, such as vertebrate neurons. To visualize the movement of RNPs in the axons of Xenopus RGCs, we labeled endogenous RNAs by blastomere injection of the fluorescently labeled uridine-5′-triphosphate (Cy3-UTP). Cy3-UTP is incorporated into RNAs during synthesis, including rRNAs and mRNAs, allowing subsequent visualization of fluorescent Cy3-RNA granules in RGC axons of cultured embryonic eyes (Figure 1A). Single-particle tracking analysis revealed the presence of static or oscillatory (Figure 1B1), slow-moving (Figure 1B2), and fast-moving (Figure 1B3) states of RNA granule movements, with occasional switches from one state to another. Most Cy3-RNA granules displayed static or oscillatory motions over 2 min (Figure 1C), similar to single mRNAs in dendrites. A slight bias toward anterograde compared with retrograde transport (Figure 1C) is consistent with the directional movements of Neurofilament-Light (Nefl) mRNA in axons. The speed distribution of transported RNA granules could be resolved into two classes of anterograde or retrograde transport: slow and fast (Figure 1D), as observed for endogenous β-actin mRNA in axons. The average speed of individual Cy3-RNA granules was negatively correlated with their signal intensities, suggesting that slow-moving or static granules can carry a larger RNA cargo load (Figure 1E).

5To find out whether RNA granules in axons are associated with endosomes, we used GFP-Rab5a or GFP-Rab7a, which label early and late endosomes, respectively (; Figures S1A–S1D). Both exhibited typically rounded profiles, with occasional variations in size and shape indicative of different endosomal maturation and/or fusion states. By expressing GFP-Rab5a and GFP-Rab7a along with Cy3-labeled RNAs, we found that static, oscillatory, and moving Cy3-RNA granules were often associated with both early and late endosome reporters in axons (Figures 1F and 1G; Video S1). RNA granules appeared to associate and dissociate from endosomes (Figure 1G), a characteristic similar to that observed for RNPs in fungal hyphae. Cy3-RNA granules were preferentially associated with GFP-Rab7a (40% ± 3%, n = 298 RNA granules in 23 axons) compared with GFP-Rab5a (24% ± 1%, n = 347 RNA granules in 24 axons). Despite the abundance of endoplasmic reticula (ERs) throughout the axon, the levels of co-labeling between Rab5a or 7a endosomes, and ERs was comparatively low (Figures S1C and S1D), and dynamic co-imaging showed that Cy3-RNA granules moved independently of the ER (Figure S1E). Static or oscillatory movements were most prevalent among RNA granules associated with Rab5a or Rab7a endosomes (Figure 1H). Furthermore, RNA granules associated with GFP-Rab5a or GFP-Rab7a endosomes moved significantly slower than isolated RNA granules in the same axons (Figure 1I). The average pixel intensity and size of endosome-associated Cy3-RNA granules were higher and larger than those of isolated Cy3-RNA granules, suggesting the presence of more RNA (Figures 1J and 1K).

6We next asked whether Rab activity plays a role in global RNA granule trafficking in RGC axons. Expression of constitutively active (CA) or dominant-negative (DN) forms of Rab5a, but not Rab7a, induced a visible reduction in axonal growth, as reported previously. However, expression of mutant forms of either Rab5a or Rab7a did not affect the RNA granule speed distribution profile or directionality compared with controls (Figures S2A–S2D).

7Collectively, these results suggest that, although RNA granules are mostly transported and distributed in axons in an endosome-independent manner, they frequently associate with both early and late endosomes along the axon (Figure S2E).

8Previous mass spectrometry analysis of endosomal composition in neurons revealed the presence of RBPs and translation machinery. We therefore performed immunocytochemistry against known RNP components in RGC axons. Rab5, Rab7, and the late endosome marker LAMP-1 partially colocalized with RBPs such as Vg1RBP (zipcode-binding protein-1) or the Xenopus RBP Fragile X-related (FXR) as well as ribosomal proteins in axon shafts (Figures 2A and 2B).

9To study these associations in more detail, we analyzed ribosomal proteins, Vg1RBP, and its β-actin mRNA cargo using live imaging. Ribosomal proteins, Vg1RBP, and β-actin mRNA each associate with GFP-tagged Rab5a and Rab7a proteins, as seen using immunoprecipitation (IP) assays (Figures S3A and S3B), and super-resolution microscopy revealed that ribosomal proteins closely associated with GFP-Rab5a and GFP-Rab7a endosomes (Figure S3C). The ribosomal protein RPS3A-GFP exhibited a mostly diffuse pattern of fluorescence interspersed with foci of higher intensity along the axon shaft. Live co-imaging of red fluorescent protein (RFP)-Rab5a, RFP-Rab7a, LAMP-1-RFP, and RPS3A-GFP revealed an association between the ribosomal and endosomal markers (Figure 2C; Video S2) that exhibited coordinated motions, mostly oscillatory or static, that persisted throughout the 1-min imaging period (Figure 2D). Similar results were obtained by co-expressing another ribosomal protein, RPS4X-GFP, with RFP-Rab5a or RFP-Rab7a (Figure S3D). Although the majority of the GFP-Vg1RBP signal was not co-transported with endosomes, we found that a fraction of GFP-Vg1RBP also joined slow-moving endosomes over the 1-min recording periods (Figures 2E and 2F; Video S3) and displayed static or oscillatory motions (Figure 2G). To track endogenous β-actin mRNA directly, we used Cy3-molecular beacons (MBs). MBs are hairpin-shaped oligonucleotide probes with a fluorophore and a quencher that separate upon hybridization to target sequences, allowing visualization of endogenous transcripts. MB tracking revealed that approximately 25% and 35% of β-actin mRNA granules were, indeed, associated with GFP-Rab5a and GFP-Rab7a (Figures 2H and 2J), of which most were static or oscillatory (Figure 2I). Together, these results indicate that ribosomes, RBPs, and mRNAs frequently assemble together with endosomes in axons.

10Static RNA granules labeled with fluorescent UTPs are hotspots of newly synthesized proteins in axon terminals in vivo. Our findings above raise the possibility that these hotspots are on endosomes. To test this idea, we first investigated whether de novo protein synthesis in RGC axons is affected by altered Rab function. Axon-only cultures (after soma removal) were pulse-labeled with a low concentration of puromycin (2 μM), a structural analog of aminoacyl-tRNAs that incorporates into the C termini of nascent polypeptide chains. Puromycylated peptides can be recognized by anti-puromycin antibodies, allowing quantification of local protein synthesis. We found a significant reduction in puromycin signal in growth cones expressing CA mutant GFP-Rab7aQ67L (0.74 ± 0.03) or DN mutant GFP-Rab7aT22N (0.81 ± 0.04) but not wild-type GFP-Rab7a, wild-type GFP-Rab5a, CA mutant GFP-Rab5aQ80L, or DN mutant GFP-Rab5aS35N (Figures 3A and 3B). Next, we acutely perturbed endosomal sorting pharmacologically (Figure 3C). Inhibition of endocytosis with dynasore, a small GTPase inhibitor targeting dynamin, for 20 min did not affect the puromycin signal, whereas the late endosomal acidification inhibitor chloroquine (CHQ) significantly attenuated it. The reduction in puromycin signal following CHQ treatment or with Rab7a mutant expression in axons was not as severe as that following treatment with the protein synthesis inhibitor cycloheximide (CHX) (Figure 3C), indicating that axonal protein synthesis is not exclusively endosome-dependent. We also saw a clear enrichment of the puromycin signal associated with GFP-Rab7a but not GFP-Rab5a endosomes (Figures 3D–3F), which was abolished by CHX treatment (Figure 3F). This enriched puromycin signal was also reduced by GFP-Rab7aQ67L or GFP-Rab7aT22N (Figure 3F) or application of CHQ (Figure 3F), confirming that both Rab7a activity and late endosome maturation are essential for late endosome-sited translation.

11Axonal protein synthesis is regulated by target of rapamycin complex 1 (TORC1) cap-dependent translation initiation. Application of the TOR inhibitor PP242 significantly reduced late endosome-sited protein synthesis (Figure 3F). We found that TOR co-precipitated with endogenous Rab7 (Figure S4A) and GFP-Rab7a in brain lysates (Figure S4B) and colocalized with GFP-Rab7a in RGC axons (Figure S4C). We then measured phosphorylated ribosomal protein S6 (p-S6), a marker of TOR-mediated translation activity, in Rab7 mutant axons. Expression of GFP-Rab7aQ67L or GFP-Rab7aT22N mutants led to a decrease in p-S6 levels in axons compared with GFP-Rab7a and GFP (Figure S4D) without affecting the upstream TOR activation pathways (Figures S4E and S4F).

12Mitochondrial respiration helps to sustain protein synthesis, and, in axons, focal translation hotspots have been reported to correlate with nearby mitochondria. We observed that Rab7a endosomes are frequently in contact with mitochondria in axons (Figures S5A and S5B). Live imaging showed that they often pause when they encounter mitochondria and form apparent contacts that persist for over 2 min (Figures S5C and S5E). Although close associations were also seen between Rab5a endosomes and mitochondria, these associations were brief and rarely persisted over prolonged periods (Figures S5D–S5F). In agreement with observations in HeLa cells, we found that contacts between constitutively active Rab7a (GFP-Rab7aQ67L) endosomes and mitochondria in axons remained for a longer time than those between wild-type Rab7a endosomes and mitochondria (Figures S5D–S5F).

13To explore whether late endosomes in association with mitochondria coincide with translational hotspots, we first investigated the spatial relationship between RNA granules, mitochondria, and late endosomes (Figure 3G; Video S4). Approximately 35% of GFP-Rab7a endosomes were found adjacent to mitochondria, and 80% of these were associated with RNA granules (Figure 3H). We then asked whether newly synthesized proteins are found at these sites (Figures 3I and 3J). Indeed, 76% of Rab7a endosomes in close proximity to mitochondria had an enriched puromycin signal (Figure 3K), suggesting that late endosomes adjacent to mitochondria are sites of mRNA translation. We also noted that 44% of late endosomes with an enriched puromycin signal were not associated with mitochondria, indicating that not all endosome-associated translation occurs in the presence of mitochondria.

14The observed spatial coincidence of mitochondria, late endosomes, and nascent proteins raises the possibility that endosomes are sites for the translation of mRNAs encoding proteins that regulate or maintain mitochondrial function. To test this idea, we first focused on the RBP SFPQ (splicing factor proline- and glutamine-rich), which coordinates the axonal transport of mRNAs critical for mitochondria and is enriched near mitochondria and ribosomes in axons. In addition to the previously identified RBPs, we found that SFPQ co-precipitated with endogenous Rab7 (Figure 4A) and with GFP-Rab7a (Figure 4B) in brain lysates. In axons, 47% of GFP-Rab7a endosomes colocalized with SFPQ (n = 137 vesicles in 80 axons) (Figure 4C).

15Time-lapse imaging showed an association between Cy5-labeled laminB2 mRNA (lmnb2), one of the SFPQ-regulated mRNAs important for mitochondrial integrity and axon survival, and GFP-Rab7a endosomes in axons (Figure 4D). To investigate whether the Rab7a endosome-associated lmnb2 mRNA is translationally active, we used live single-molecule nascent polypeptide imaging based on the SunTag fluorescent tagging system to visualize axonal Lamin B2 (LB2) protein synthesis in real time (Figure 4E;,, ). As the nascent peptide is synthesized, super-folder GFP-tagged single-chain antibody fragments (scFv-sfGFP) rapidly bind to the 24 SunTag epitopes fused to the N-terminal of LB2, visible as a bright fluorescent spot. Cy5-labeled SunTag-lmnb2 mRNA was electroporated into eye primordia of embryos expressing RFP-Rab7a and scFv-sfGFP. We observed a striking colocalization between the three signals: RFP-Rab7a endosomes, Cy5-SunTag-lmnb2 mRNA, and sfGFP-labeled newly synthesized SunTag-LB2 proteins in axons (Figure 4F; Video S5). As expected, the proportion of actively translating mRNAs in axons or on late endosomes, indicated by the percentage of Cy5-SunTag-lmnb2 colocalizing with SunTag-LB2 spots, was decreased by exposing the axons to a high concentration of puromycin, which blocks protein synthesis and induces the release of nascent polypeptides (Figures 4G and 4H). Interestingly, the percentage of Rab7a endosomes associated with SunTag-lmnb2 mRNA remained unchanged (Figure 4I). The same effect was observed following application of CHQ (Figures 4G-I) suggesting that perturbing late endosome maturation impairs mRNA translation, not mRNA binding.

16To see whether endogenous lmnb2 mRNA could be observed on endosomes, we first used co-immunoprecipitation (coIP) followed by qRT-PCR and found that endogenous lmnb2 mRNA was associated with Rab7a endosomes (Figure 4J). We then visualized the synthesis of native LB2 using puromycylation coupled with a proximity ligation assay (Puro-PLA) and found the LB2 Puro-PLA signal in close proximity to Rab7a endosomes (Figure 4K). This was further confirmed by super-resolution microscopy, which revealed LB2 Puro-PLA puncta decorating Rab7a endosomes (Figure 4L). Application of CHQ, CHX, and PP242 significantly reduced the amount of LB2 Puro-PLA puncta in axons (Figure 4M), indicative of protein synthesis inhibition. To extend these findings to another mitochondrion-related protein, we examined the axonally synthesized voltage-dependent anion-selective channel protein 2 (VDAC2), which is involved in exchanging solutes across the outer mitochondrial membrane. We found that vdac2 mRNA also co-precipitated with GFP-Rab7a (Figure 4N). We then confirmed that VDAC2 was locally synthesized by Puro-PLA and that CHQ, CHX, and PP242 reduced the signal in axons (Figure 4O). Taken together, these results show that late endosome-coupled translation can supply nascent proteins that have a role in maintaining axonal mitochondrial function.

17The autosomal-dominant neurological disorder CMT2B is associated with missense mutations of four amino acids in the human rab7a gene. This rare neuropathy primarily affects the peripheral sensorimotor systems, leading to distal sensory loss, muscle weakness, and atrophy. The mechanism behind how these mutations induce axonopathy remains elusive. We wondered whether our findings of late endosome-associated axonal translation, especially of mRNAs essential for mitochondrial and axonal integrity, could be relevant to understanding CMT2B. Because the amino acid sequence of Xenopus Rab7a is 96% identical to human Rab7a, including the four residues identified as mutated in CMT2B patients (L129F, K157N, N161T/I, and V162M), we generated four GFP-tagged pathological mutants (Rab7aK157N, Rab7aL129F, Rab7aV162M, and Rab7aN161T) and expressed them in the Xenopus RGCs. In distal RGC axons expressing CMT2B-linked mutants, discrete GFP-positive puncta were clearly visible with each mutant, resembling patterns observed in wild-type GFP-Rab7a and CA GFP-Rab7aQ67L, in contrast to the more diffuse signal observed for the DN GFP-Rab7aT22N (Figure 5A). Live imaging revealed that expression of CMT2B mutants affects the direction, frequency of pausing, and average speed of LysoTracker-positive vesicles in RGC axons (Figures 5B–5F), suggesting conserved dominant effects of these mutants on late endosomal trafficking in our system.

18Severe defects in RGC axonal projections in vivo were also noted in DiI labeling experiments at stage 45, when most of these axons have established synaptic connections in the optic tectum of the midbrain. The majority of embryos expressing Rab7a disease mutants had fewer axons in the optic tract and the optic tectum (Figures 5G–5I). To test whether this effect on axons was autonomous, we performed mRNA injections into only one of the two dorsal blastomeres at the four-cell stage, leading to embryos in which expression was restricted to one-half of the CNS (Figure 5J). As RGC axons in Xenopus cross at the optic chiasm, contralaterally projecting mutant axons navigate through wild-type tissue. Although less severe, RGCs axons expressing GFP-Rab7aV162M or GFP-Rab7aN161T in the wild-type brain also displayed less dense projections at stage 45 (Figures 5K and 5L). To distinguish whether the defects at stage 45 optic projections were the result of a failure to grow initially or degeneration after growth, we examined the projections 3–4 days earlier, at stage 39. No obvious axon projection phenotypes were observed in mutant-expressing embryos at stage 39 (Figures 5M and 5N), suggesting that Rab7a-CMT2B mutants affect axonal maintenance after their initial growth.

19The loss of axonal integrity induced by CMT2B mutants could be preceded by impaired intra-axonal translation. Indeed, puromycin labeling revealed a decrease in nascent protein synthesis in growth cones of somaless RGC axons from stage 33/34 embryos expressing each of the four CMT2B mutants compared with GFP-Rab7a or the GFP control (Figures 6A and 6B). The translation marker p-S6 in growth cones was also reduced by all four of the CMT2B-related Rab7a mutants (Figure 6C). Finally, the percentage of Rab7a CMT2B mutant endosomal puncta colocalizing with enriched puromycin signal was significantly decreased compared with wild-type GFP-Rab7a endosomes (Figure 6D).

20To test whether the synthesis of proteins needed for mitochondrial function was affected in axons expressing CMT2B mutants, we first used the SunTag method and found that the proportion of actively translating SunTag-lmnb2 mRNAs was significantly reduced in RFP-Rab7aV162M- or RFP-Rab7aN161T-expressing axons (Figure 6E). Moreover, there was a significant reduction in newly synthesized SunTag-LB2 proteins on mutant-associated endosomes compared with wild-type RFP-Rab7a endosomes (Figure 6F). Interestingly, the percentage of wild-type or mutant endosomes associated with Cy5-labeled mRNA remained constant (Figure 6G), suggesting that Rab7a disease mutations do not alter the association of lmnb2 mRNAs with late endosomes but, rather, diminish their translation.

21Next, we asked whether the synthesis of native LB2 and VDAC2 proteins was affected by expressing CMT2B-related Rab7a mutants. Overexpression of GFP-Rab7a led to an approximately 20% decrease in nascent axonal LB2, as measured by LB2 Puro-PLA (Figures 6H and 6I), suggesting that LB2 local synthesis is sensitive to altered Rab7a levels in axons. Expression of each of the four Rab7a CMT2B mutants, however, caused a significantly larger drop (an approximately 35% decrease) in new LB2 synthesis (Figures 6H and 6I). Three of four mutants, but not wild-type GFP-Rab7a, resulted in a reduction in VDAC2 Puro-PLA puncta (Figure 6J). Together, these results suggest that CMT2B Rab7a mutations compromise axonal translation of mRNAs, including those encoding proteins essential for mitochondrial integrity.

22Because CMT2B-associated Rab7a mutants downregulate axonal translation of mitochondrial proteins, we wondered whether mutant-expressing axons exhibit abnormal mitochondrial phenotypes. To address this question, we first analyzed mitochondrial morphology in RGC axons expressing each of the four CMT2B Rab7a mutants. MitoTracker-labeled mitochondria in CMT2B Rab7a mutant-expressing axons exhibited highly elongated profiles (Figure 7A) compared with GFP-Rab7a or GFP controls (Figure 7B), whereas the density of mitochondria remained unchanged (Figure 7C), consistent with the mitochondrial phenotypes seen when axonal LB2 synthesis was inhibited. Defective axonal mitochondrial trafficking is a sign of altered mitochondrial physiology, so we analyzed mitochondrial dynamics in axons expressing Rab7aV162M during 5-min time-lapse imaging intervals. We found a decrease in anterograde mitochondrial transport in CMT2B mutant-expressing axons (Figure 7D) and an increase in retrograde transport (Figure 7D) compared with GFP-Rab7a or GFP control axons. Finally, we measured the mitochondrial membrane potential (ΔΨm), which drives ATP production, by quantifying the accumulation of the cationic fluorescent probe tetramethylrhodamine methyl ester (TMRM) in mitochondria along axons and found a significant ΔΨm reduction in axons expressing Rab7aV162M compared with the GFP control or wild-type GFP-Rab7a (Figure 7E).

23Our results support a model in which Rab7a endosomes are sites for mRNA translation in axons, generating focal hotspots of local protein synthesis often associated with mitochondria (Figure 7F). Moreover, we found that translation of nucleus-encoded mitochondrial mRNAs essential for mitochondrial function occurs at these hotspots.

24Axonal mRNA localization relies on specific association with RBPs and targeted transport through direct or indirect binding with cytoskeletal motor proteins. Our live-imaging analyses of RNA granules, mRNAs, and RBPs collectively support this view because they provide evidence for directed transport of RNPs in RGC axons, mostly independent of endosomal trafficking. However, we also found that around a quarter of oscillatory and slow-moving RNPs are associated with endosomes in axons. How RNPs become tethered to the endosomes remains to be determined. A lipid-binding adaptor protein has been proposed as the mechanism for mRNA recruitment on endosomes in fungal hyphae. This mechanism may be conserved because several different RBPs (SFPQ, Vg1RBP, and FXR) were found to be part of the endosome-associated complex in our system.

25It has been suggested that ribosomes assemble on endosome-associated mRNAs to activate translation, and we found ribosomal proteins associated with late endosomes in axons. In addition, the enrichment of newly synthesized proteins at Rab7a endosomes indicates that late endosomes are hotspots of local protein synthesis. Rab7a function is required for axonal translation because both CA and DN Rab7a mutants caused downregulation of translation. Previous work has shown that the endosomal sorting complex required for transport II (ESCRT-II) colocalizes with Rab7 in RGC axons, and ESCRT-II knockdown causes a similar decrease (∼20%) in RGC axonal protein synthesis, suggesting a potential role of ESCRT-II in this process. Not all intra-axonal translation requires endosomal function; the protein synthesis inhibitor CHX decreases axonal protein synthesis more severely (∼50%). Our results also showed the presence of mRNAs and ribosomes on early Rab5a endosomes, but neither expression of Rab5a mutants nor the endocytosis inhibitor dynasore revealed significant effects on constitutive axonal protein synthesis. Future work will be needed to determine why late but not early endosomes seem to regulate axonal translation in this system.

26Translation hotspots have previously been reported in both axons and dendrites. In axons, hotspots correlate with the presence of mitochondria. In accordance with these observations, we found that hotspots of late endosome-associated translation were often in close proximity to mitochondria. Strikingly, live imaging revealed that Rab7a endosomes remained in contact with mitochondria for prolonged periods. A direct association between late endosomes or lysosomes and mitochondria has been reported in HeLa cells and correlates with mitochondrial fission events. However, no mitochondrial fission event was observed in RGC axons when late endosomes docked on mitochondria, suggesting that other molecular activities may also be involved in these interactions. Mitochondrion-linked mRNAs represent one of the main categories of translating mRNAs in adult axon terminals in the mouse, and their local translation is essential for the maintenance of mitochondrial function. We found that late endosomes are sites of axonal translation of mRNAs that are essential for mitochondria. Indeed, the RBP SFPQ, which orchestrates a mitochondrion-related RNA regulon, and its cargo lmnb2 mRNA were found on late endosomes in axons. Moreover, we showed that Rab7a endosomes are sites for the local synthesis of LB2, which is key to mitochondrial and axonal integrity. It is not yet clear whether the association with mitochondria can stimulate late endosomal-sited translation, and identification of a tethering complex may be essential to answer this question. In addition, the observed close association of the ER with RNA granules and late endosomes in RGC axons opens the possibility of its involvement in late endosomal trafficking and/or translation-associated activity. It is also of interest to note that, although mitochondrial mRNAs represent only 5% of the total RGC axonal translatome, global axonal protein synthesis decreases by 20% when Rab7a function or late endosome maturation is perturbed. This suggests that endosome-associated translation is involved in other local physiological processes in axons, and our finding that β-actin mRNAs also associate with endosomes supports this idea.

27The inherited peripheral neuropathy CMT2B is caused by any one of the five rab7a missense mutations (e.g., L129F, K157N, N161T/I, and V162M). Despite Rab7a being a ubiquitous protein, axons of peripheral neurons are particularly susceptible to CMT2B-associated mutations. Different molecular mechanisms have been proposed to explain this. rab7a CMT2B-associated mutations show perturbed axonal transport of late endosomes, an effect we also observed in our system. These disrupted endosomal dynamics are correlated with altered signaling responses to neurotrophic factors in axons, in particular nerve growth factor (NGF). Another explanation is that rab7a mutations disrupt the regulation of effectors that function specifically in the peripheral nervous system, such as the intermediate filament peripherin, which has been shown to interact directly with Rab7a. In this study, we propose an additional mechanism that may contribute to CMT2B disease; i.e., that Rab7a mutants affect local translation. Local translation is thought to be particularly important in the distal portions of long axons, such as sensory and motor neurons, because it supplies new proteins to meet local demand far from the soma. In this regard, it is of note that the axonal phenotypes observed in CMT2B-expressing peripheral sensory axons, and retinotectal axons in this study, are typical of the axonal phenotypes seen when subcellular RNA-based mechanisms are dysregulated. We found compromised local protein synthesis in CMT2B mutant-expressing axons and, in particular, of known axonal survival factors, such as LB2. The fact that CMT2B mutants did not affect the association of lmnb2 mRNAs with late endosomes but did reduce protein synthesis suggests that these Rab7a disease mutants affect the efficiency of endosome-sited translation. Further work will be required to understand the exact mechanistic role of Rab7a in this process and how much the reduced local translation contributes to the axonal and mitochondrial phenotypes seen in CMT2B mutant-expressing axons. In view of the emerging functional interactions between endosomes and mitochondria, it will also be of interest in the future to find out whether other aspects of the cross-talk between these two organelles are modulated by Rab7a functions in axons and whether these are perturbed in CMT2B disease.

28Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Christine E. Holt (ceh33@cam.ac.uk).

29Xenopus laevis eggs were fertilized in vitro and embryos were raised in 0.1x Modified Barth’s Saline (MBS; 8.8mM NaCl, 0.1 mM KCl, 0.24mM NaHCO3, 0.1 mM HEPES, 82μM MgSO4, 33μM Ca(NO3)2, 41μM CaCl2) at 14-20°C and staged according to the tables of Nieuwkoop and Faber. All animal experiments were approved by the University of Cambridge Ethical Review Committee in compliance with the University of Cambridge Animal Welfare Policy. This research has been regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB).

30Eye primordia were dissected from Tricaine Methanesulfonate (MS222) (Sigma-Aldrich) anesthetized embryos of either sex at stage 35/36 and cultured on 10μg/ml poly-L-lysine (Sigma-Aldrich)- and 10μg/ml laminin (Sigma-Aldrich)-coated glass bottom dishes (MatTek) in 60% L-15 medium (ThermoFisher), 1x Antibiotic-Antimycotic (ThermoFisher) at 20°C for 24-48 hours. 10-20 eye primordia (from 5-10 embryos) were cultured per dish and, typically, 2-3 dishes were used per experimental condition for each biological replicate.

31GFP- or RFP-tagged Xenopus Rab constructs GFP-Rab5a, RFP-Rab5a, GFP-Rab7a, RFP-Rab7a, constitutively active mutant GFP-Rab7aQ67L, dominant negative mutant GFP-Rab7aT22N in pCS2+ vector were previously reported. Point mutations of GFP-Rab5a to obtain the constitutively active mutant GFP-Rab5aQ80L and dominant negative mutant GFP-Rab5aS35N, and GFP-Rab7a or RFP-Rab7a to obtain the four CMT2B-associated mutants Rab7aK157N, Rab7aL129F, Rab7aV162M and Rab7aN161T were performed by site-directed mutagenesis using QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies). Human and Xenopus Rab5a or Rab7a protein sequences were aligned to identify the amino acids to be mutated. Lamp1-RFP (Addgene) was cloned into pCS2+ vector. Xenopus rps3a and rps4x sequences were obtained by PCR from a Xenopus laevis cDNA library synthesized using SuperScript III First-Strand Synthesis System (ThermoFisher Scientific), and subsequently cloned into the N-terminal of GFP in pCS2+ vectors to construct Rps3a-GFP and Rps4x-GFP plasmids. Xenopus Vg1RBP was obtained from pET21d-Vg1RBP-GFP by PCR and cloned into a GFP-containing pCS2+ vector, which has been previously described. Mitochondria-targeted GFP (Mito-GFP) was a gift from Michael Coleman (Department of Clinical Neuroscience, University of Cambridge, UK). Xenopus laminb2 (lmnb2) cDNA sequence including the 5′ and 3′ untranslated regions (UTRs) was obtained from IMAGE clone (Source Bioscience SB5157193) by PCR and cloned into pCS2+ vector for Cy5-labeled mRNA synthesis. 24 SunTag tandem repeats were obtained from pcDNA4TO-24xGCN4_v4-kif18b-24xPP7 (Addgene) by PCR and cloned into pCS2+ vector. lmnb2 5′UTR was cloned into 5′ end of the SunTag repeats, while lmnb2 coding sequence (CDS) and 3′UTR were cloned into 3′ end of the SunTag repeats, generating the lmnb2 5′UTR-24xSunTag-lmnb2 CDS-lmnb2 3′UTR construct, which was used to generate Cy5-labeled SunTag-lmnb2 mRNA. scFv-sfGFP sequence obtained from pHR-scFv-GCN4-sfGFP-GB1-dWPRE (Addgene) by PCR was cloned into pCS2+ vector. The E.coli DH5α strain was used for all plasmid amplification steps. Capped RNAs were in vitro transcribed using mMessage mMachine SP6 Transcription Kit (ThermoFisher Scientific). In case of Cy5-labeled mRNA synthesis, 1μl of Cy5-UTP (PerkinElmer) was added to the reaction mixture during in vitro transcription.

32DNA and RNA were microinjected into both of the dorsal blastomeres at four- or eight-cell stage as previously described. Embryos were de-jellied with 2% cysteine (Sigma-Aldrich) in 1X MBS (pH 8), washed 3 times in 0.1X MBS and aligned on a grid in 4% Ficoll (Sigma-Aldrich) in 0.1X MBS, 1% penicillin (100 U/ml), streptomycin (100 μg/ml) and fungizone 0.25 μg/ml (Antibiotic-Antimycotic, GIBCO). Injections of 5 nL of volume were performed using glass capillary needles (outer diameter: 1.0 mm; inner diameter: 0.5 mm, Harvard Apparatus) and a microinjector (Picospritzer, General Valve). Cy3-UTP (PerkinElmer) was injected at 100μM. RNA coding for GFP and GFP- or RFP-tagged wild-type and mutant Rab5a were injected at a concentration of 100ng/μl. RNA coding for GFP and GFP- or RFP-tagged wild-type and mutant Rab7a were injected at a concentration of 200ng/μl. Mito-GFP and GFP-Vg1RBP DNA plasmids were injected at a concentration of 25ng/μl. Lamp1-RFP DNA plasmid was injected at a concentration of 50ng/μl.

33DNA plasmids [1μg/μl] or in vitro transcribed RNAs [1μg/μl] were introduced into Xenopus eye primordia of stage 26-30 embryos by electroporation. Anaesthetized embryos were transferred to a “†” shaped chamber with the head positioned at the cross intersection and platinum electrodes positioned on either side, in the transverse channel. A borosilicate glass capillary needle (outer diameter: 1mm; inner diameter: 0.78mm, Harvard Apparatus) containing the DNA or RNA solution was inserted into the eye and a volume of 40nl was injected. The capillary was withdrawn from the eye post-injection, immediately prior to delivering eight square wave electric pulses of 18V, with 50ms duration and 1000ms intervals. After electroporation, embryos were transferred to 0.1x MBS for recovery and electroporated eyes were cultured at stage 35/36. For live-imaging experiments, Rps3a-GFP or Rps4x-GFP DNA plasmids were electroporated into RFP-Rab5a or RFP-Rab7a RNA- or Lamp1-RFP DNA-injected embryos. In vitro transcribed Cy5-labeled lmnb2 mRNA was electroporated into GFP-Rab5a or GFP-Rab7 RNA-injected embryos.

34Cultured axons were imaged under a Perkin Elmer Spinning Disk UltraVIEW ERS, Olympus IX81 inverted microscope with a 60x 1.4NA silicone oil objective for 2 to 5 minutes with exposure times set between 300-500ms. Volocity (Perkin Elmer) was used for manual tracking of Rab7a or Rab5aendosomes associated with either Cy3-RNA labeled granules or GFP-Vg1RBP. Associated signals were defined by their association/co-movement for > 4 s. A granule was defined as static or oscillatory if no directional transport further than 2μm within 60 s was observed. To visualize ER, retinal cultures were incubated with 1μM ER-Tracker Red (ThermoFisher Scientific) for 15 minutes and washed 7 times with culture medium. To visualize acidic endosomes, retinal cultures were incubated with 50nM Lyso-Tracker Red (ThermoFisher Scientific) for 30 minutes and washed 7 times with culture medium. Movies of Rab7a endosomes and mitochondria were taken for 5 minutes directly after the cultures were incubated for 20 minutes in 50nM MitoTracker red CMXRos (ThermoFisher Scientific) dissolved in DMSO and washed 7 times with culture medium. To quantify mitochondria membrane potential, retinal cultures were incubated with 20nM tetramethylrhodamine methyl ester (TMRM) (ThermoFisher Scientific) for 20 minutes, followed by 4 washes and immediate image acquisition of single frames with constant laser power and exposure time. Object detection analysis on Volocity was then used to quantify the mean intensity of the TMRM signal in each mitochondrion present in 50μm of isolated axon shafts expressing the indicated constructs, excluding the last 20μm distal portion of the axon. The mitochondria outlines were then placed in an adjacent area clear of mitochondria to record the background fluorescent intensity. This reading was subtracted yielding the background-corrected intensity.

35Particle movements were extracted from raw data movies with the plusTipTracker package. We used the Watershed-based option in the Detection Settings for all movies. To detect individual features (granules), movies were denoised with 2 Gaussian Kernels of size 1 pixel and 3 pixels respectively, corresponding to the sigma 1 & 2 in the Detection Settings of the plusTipTracker package. The K-Value (local threshold) was set to values between 2-5 depending on the background noise in the movies. The Tracking settings were held constant for all movies: Search Radius Range = 3-10, Minimum Sub-Track length = 4 frames (to discard background fluctuations), Maximum Gap Length = 3 (frames for which a particle is not detected due to overlap with another particle), Max Shrinkage Factor = 0, Maximum angle Forward/Backward = 30/0, Fluctuation Radius = 1.5. After running this analysis, several tracks (x-y coordinates) were generated for each movie.

36The tracks representing Cy3-RNA granule or LysoTracker vesicle movements in each movie were then post processed in a self-written MATLAB script to determine their motion types. First, a segmented line was drawn along the axon toward the anterograde direction. Then each (average) track direction was calculated. If the average direction points within 162 degrees of the forward/anterograde direction of the drawn axon, the track is classified as anterograde and vice versa for retrograde. Furthermore, tracks that displace less than 2μm from their origin were classified as oscillatory. To calculate the Cy3-RNA track/granule intensity a background region was drawn by hand and the average pixel intensity in this region subtracted from the average intensity of the granule for each track (Individual granule intensities are calculated in the detection part of the plusTipTracker script). The average track velocity was calculated as the average of all frame-to-frame velocities.

37Images of fluorescently labeled RNA granules, Rab5a endosomes, Rab7a endosomes, or mitochondria were analyzed using custom-written scripts within the MATLAB environment. All grayscale fluorescent images were binarized using Otsu’s thresholding method. The binary images were segmented using a modified watershed method to obtain masks of labeled granules from each fluorescent channel. The resulting images were then visually inspected and the threshold was adjusted where granules were misidentified.

38The masks from the Cy3-RNA channel were compared spatially to each of the other three and the granules that overlapped with GFP-Rab5a, GFP-Rab7a or mitochondria, by at least one pixel, were identified. The Cy3-RNA granules were thus grouped as in contact or not in contact with the other component. The area of each granule was defined as the total number of pixels of its mask and was normalized to the smallest and largest granule areas within one image (typically of a single axon). Since the mask areas of brighter granules tended to be larger, we calculated the mean intensities of the brightest pixels within each mask to remove any possible dampening of the signal due to mask size. The number of brightest pixels to be included was separately determined for each of Cy3-RNA granule pairs and set as half of the average area of Cy3-RNA granule masks in that pair. Pixel intensities of the granules were obtained from the original grayscale image and normalized such that the minimum value would equal the binarization threshold and the maximum value was that of the brightest pixel within the corresponding image’s masks.

39The masks from RFP-Rab7a were compared to those from mitochondria to find the distances between objects in each image. Where an overlap between two objects from each channel existed, the distance between those objects was set to zero. Otherwise, the shortest distance between an object and its nearest neighbor in the other channel was calculated. These shortest distances for all objects were collated and analyzed. To create a randomized distribution of endosomes and mitochondria for comparison, the mean and standard deviations of their mask radii, number of particles, and axon dimensions were calculated. Normal distributions based on these mean and standard deviations were created for each of the above variables. Random numbers of each particle type were drawn from normal distribution of particle numbers. Each particle was a circle of random radius drawn from the respective normal distribution. Particles were placed in axons whose width and length were drawn from corresponding distributions, as well. The distance between these randomly selected particles were measured and analyzed as described above.

40For GFP immunoprecipitations, GFP, GFP-Rab5a or GFP-Rab7a expressing Xenopus brains were dissected from stage 35/36 embryos (70 embryos/condition) and homogenized in lysis buffer (20mM Tris, 100mM NaCl, 10mM MgCl2, 0.25% NP40, 10% Glycerol, 100 μg/ml CHX (Sigma-Aldrich)) supplemented with EDTA-free protease inhibitor cocktail (Roche), phosphatase inhibitor (Pierce) and 100U/ml SUPERase In RNase inhibitor (Ambion) for 5 minutes on ice. Following centrifugation for 5 minutes at 1000 x g, the supernatant was collected. For immunoprecipitation, the protein extracts were incubated for 30 minutes at 4°C with beads coupled to an anti-GFP antibody (ab1218, Abcam). Beads were washed four times with lysis buffer and samples were either eluted from the beads by adding RLT buffer (QIAGEN) and vortexing for 2 minutes for subsequent RNA isolation or eluted with 1x SDS sample buffer for 5 minutes at 95°C for protein and subsequent western blot analysis as described below. For endogenous immunoprecipitations, brains from stage 35/36 embryos (70 embryos/condition) were dissected and lysed in lysis buffer as described above. Lysates were then incubated with beads coupled to anti-Rab7 (ab77993, Abcam) or IgG control (ab37415, Abcam) antibodies for 30 minutes at 4°C. Beads were washed three times with lysis buffer and samples were eluted from the beads with 1x SDS sample buffer for 5 minutes at 95°C for protein and subsequent western blot analysis.

41RNA was isolated from eluted samples using the RNeasy Mini kit (QIAGEN) and reverse transcribed into cDNA using random hexamers and the SuperScript III First-Strand Synthesis System (ThermoFisher Scientific). The cDNA was used to prepare triplicate reactions for qPCR according to manufacturer’s instructions (QuantiTect SYBR Green PCR kit, QIAGEN) and plates were centrifuged shortly and run on a LightCycler 480 (Roche) using the following PCR conditions: denaturation for 15 s at 94°C; annealing for 30 s at 60°C; extension for 30 s at 72°C. The levels for each condition were corrected with their own input. The following primers were used for qPCR: for β-actin, 5′-CCAGAAGAACACCCAGTGCT-3′ and 5′-CAGGGACAACACAGCTTGGA-3′; for lmnb2, 5′-GCAAGTGAAGATGTACAAGGAAGAA-3′ and 5′-CGTCGCTCAGTTAATTCTTCTAGGG-3′; for vdac2, 5′-ACTCGCATGCAGCCATATCT-3′ and 5′-TACCAGCGCAGAAAGTGTGA-3′.

42Proteins were resolved by SDS-PAGE on NuPage 4%–12% Bis-Tris gels (ThermoFisher Scientific) and transferred to nitrocellulose membrane (Bio-Rad). The blots were blocked in milk for 45 minutes at room temperature and then incubated with primary antibodies in milk overnight at 4°C. After washing 3 times with TBS + 0.1% Tween (TBS-T) the blots were incubated with HRP-conjugated secondary antibodies for 1 hour at RT, washed again for 3 times in TBS-T, followed by ECL-based detection (Pierce ECL plus, ThermoFisher Scientific). The following primary antibodies were used for western blot analysis: rabbit anti-Rab7 (ab137029, Abcam), rabbit anti-Rpl10A (16681-1-AP, Proteintech), mouse anti-Rps3A (ab194670, Abcam), mouse anti-FXR (gift from Dr. Khandjian), rabbit anti-SFPQ (ab38148, Abcam), rabbit anti-Vg1RBP (gift from Dr. Standart), rabbit anti-mTOR (#2983, Cell Signaling).

43Retinal cultures were fixed in 2% formaldehyde/7.5% sucrose in PBS for 20 minutes at 20°C. Antigen retrieval by steaming the fixed cultures in 0.01M sodium citrate (0.05% Tween, pH 6.0) for 10 minutes was carried out before staining for phospho-AKT (Ser473), phospho-mTOR (Ser2448) and phospho-S6 (Ser235/236). The cultures were then washed 3 times in PBS + 0.001% Triton-X-100, permeabilized for 3-5 minutes in 0.1% Triton-X-100 in PBS, or 25 minutes in 0.1% Saponin (for LAMP-1 staining), washed 3 times in PBS + 0.001% Triton-X-100 and blocked with 5% heat-inactivated goat serum in PBS for 30 minutes at 20°C. Primary antibodies were incubated overnight, followed by Alexa Fluor-conjugated secondary antibodies for 45 minutes at 20°C in dark. Cultures were mounted in FluorSave (Calbiochem) or Vectashield (Vector Laboratory) for super-resolution microscopy. Antibodies were used at the following dilutions. 1:200 rabbit anti-RPS3A (14123-1-AP, Proteintech), 1:200 mouse anti-RPS3A (ab194670, Abcam), 1:200 mouse anti-RPL19 (ab58328, Abcam), 1:200 rabbit anti-RPL24 (17082-1-AP, Proteintech), 1:400 rabbit anti-Vg1RBP (gift from Dr. Standart), 1:250 rabbit anti-LAMP-1 (ab24170, Abcam), 1:400 mouse anti-FXR (gift from Dr. Khandjian), 1:50 mouse anti-Rab5 (sc-46692, Santa Cruz), 1:300 rabbit anti-Rab7 (ab137029, Abcam), 1:50 mouse anti-GM130 (BD610822, BD Biosciences), rabbit anti-Rpl10A (16681-1-AP, Proteintech), 1:100 rabbit anti-phospho-AKT (Ser473) (9271T, Cell Signaling), 1:200 rabbit anti-phospho-mTOR (Ser2448) (ab109268, Abcam), 1:200 rabbit anti-mTOR (#2983, Cell Signaling), 1:800 rabbit anti-phospho-S6 Ribosomal Protein (Ser235/236) (#2211, Cell Signaling), 1:200 rabbit anti-SFPQ (ab38148, Abcam). Secondary antibodies: 1:500 goat anti-mouse/rabbit Alexa Fluor 488 (ThermoFisher Scientific) or 1:500 goat anti-mouse rabbit Alexa Fluor 568 (ThermoFisher Scientific), except 1:1600 goat anti-rabbit Alexa Fluor 568 for phospho-S6, 1:1000 goat anti-rabbit Alexa Fluor 568 (ThermoFisher Scientific) for SFPQ immunohistochemistry.

44To visualize endogenous β-actin mRNA in Xenopus laevis RGC axons, two molecular beacons (MBs) were designed to target predicted single stranded regions in β-actin mRNA. Briefly, the MBs were made of nuclease-resistant nucleotides which were either LNA (+N) or 2′-O-methyl RNA (oN) bases, ending with Cy3:BHQ-2™ fluorophore:quencher pair combination. Probe sequences were: MB1 Cy3-oCoGoAoCoGoCoU+CoAoGoUoU+AoGoG+AoUoUoUoUoC+AoUoGoCoGoUoCoG-BHQ2; MB2 Cy3-oGoCoGoCoAoG+GoAoA+GoCoCoAoA+GoAoUoG+GoAoUoGoCoGoC-BHQ2. MB1 and MB2 were electroporated at a concentration of 25 μM into stage 28 eye primordia together with either GFP-Rab5a or GFP-Rab7a in vitro transcribed mRNA (1 μg μl-1). Eyes were dissected and cultured when Xenopus embryos reached stage 32, and RGC axons imaged the next day. Images were acquired on an Olympus IX81 inverted microscope that was fitted with a PerkinElmer Spinning Disk UltraVIEW VoX using a 100x oil immersion objective (1.4 N.A., Nikon), and an ORCA-Flash4.0 V2 CMOS camera (Hamamatsu). Volocity 6.3.0 software (PerkinElmer) was used for acquisition. Images were acquired at a frame rate of 1.5 s for 30 s. β-actin mRNA and either GFP-Rab5a or GFP-Rab7a were classed as associated if the signal from the GFP (488nm laser line) and MB-Cy3 (561nm laser line) channels merged, and remained merged for the duration of the movie, or until one of the puncta bleached (minimum 5 frames). To avoid bias in the degree of observed association due to variable levels of either MBs or GFP-Rabs proteins in axons, which was introduced by electroporation, we analyzed colocalization within 25 μm sections of axons that contained 0.8:1.5 ratio of MB:GFP-Rabs puncta. Associated puncta were defined as “oscillatory,” if they remained within a 2 μm radius for the duration of the movie, or as “moving,” if the total distance traveled was greater than 2 μm.

45After 48 hours, axons were severed from their cell bodies and subsequently treated with 2 μM puromycin (Sigma-Aldrich) for 10 minutes. After treatment, the cultures were fixed in 2% formaldehyde/7.5% sucrose in PBS for 20 minutes at RT, washed 3 times in PBS + 0.001% Triton-X-100, permeabilized for 3-5 minutes with 0.1% Triton-X-100 and blocked in 5% heat-inactivated goat serum in PBS for 30 minutes at RT and subsequently labeled with Alexa Fluor 647-conjugated mouse anti-puromycin antibody (1:250, MABE343-AF647, Millipore) overnight. For experiments involving pharmacological treatments, axons were pre-treated with 50μM CHX (Sigma-Aldrich) for 20 minutes, 50μM dynasore (Sigma-Aldrich) for 20 minutes, 50μM Chloroquine (Sigma-Aldrich) or 2.5μM PP242 for 10 minutes before puromycin administration.

46Randomly selected non-collapsed growth cones were imaged at 60x on a Nikon Eclipse TE2000-U inverted microscope equipped with an EMCCD camera. Exposure time was kept constant and below gray-scale pixel saturation. For quantitation of fluorescence intensity, the growth cone outline was traced on the phase contrast image using Volocity (PerkinElmer), then superimposed on the fluorescent image. The software calculated the fluorescent intensity within the growth cone, giving a measurement of pixel intensity per unit area. The growth cone outline was then placed in an adjacent area clear of cellular material to record the background fluorescent intensity. This reading was subtracted from the growth cone reading, yielding the background-corrected intensity. To plot the signal intensities, pixel fluorescent intensity across a line segment was measured in ImageJ. For colocalization test, Pearson’s correlation coefficient was obtained in ImageJ by using coloc2 plugin. Kymograph were obtained in ImageJ by using Multi kymograph plugin.

47For SunTag-lmnb2 imaging, in vitro synthesized scFv-GFP and RFP-Rab7a mRNAs (200ng/μl) were injected (5nl) into each of the two dorsal blastomeres at 4-cell stage. When the embryos developed to stage 26, Cy5-labeled SunTag-lmnb2 mRNA (1μg/μl) was electroporated into the retinal primordia. When the embryos reached stage 35/36, electroporated eyes were dissected and cultured. After 24 hours, continuous imaging of distal axons expressing in vitro synthesized scFv-GFP and RFP-Rab7a mRNAs was performed under a Perkin Elmer Spinning Disk UltraVIEW ERS, Olympus IX81 inverted microscope with a 100x oil objective and an ORCA-Flash4.0 camera (Hamamatsu) for 1 minute. Exposure times were set between 50-500ms for each channel. Laser power of 488, 559 and 640nm laser line was set to 10%–30% of maximal intensity. Time-lapse images were recorded by Volocity software. In pharmacologically treated conditions, the retinal cultures were incubated with 100 μg/ml puromycin or 50μM chloroquine for 1h before imaging.

48This experiment was carried out according to the manufacturer’s protocol (Duolink Biosciences) with specific modifications. Eyes were severed from axonal explants after 48 h of outgrowth. Cultures were treated with 2 μM puromycin for 10 minutes at RT, fixed in 2% formaldehyde/7.5% sucrose in PBS for 20 minutes at 20°C, washed 3 times in PBS + 0.001% Triton X-100, permeabilized for 3-5 minutes in 0.1% Triton X-100 in PBS, washed three times in PBS, blocked with 5% heat-inactivated goat or horse serum in PBS for 30 minutes at RT and subsequently incubated with primary antibodies overnight at 4°C. Primary antibodies were diluted at 1:500 for unconjugated mouse anti-puromycin (MABE343, Millipore), 1:200 for rabbit anti-Lamin B2 (ab97513, Abcam) and 1:500 for goat anti-VDAC2 (ab37985, Abcam). Dishes were washed twice for 5 minutes with 0.002% Triton X-100 in PBS and incubated with anti-rabbit (+) or anti-goat (+) and anti-mouse PLA probes for 1 hour at 37°C, with ligase for 30 minutes at 37°C and with the polymerase mix with red fluorescence for 100 min at 37°C. The samples were subsequently mounted with the mounting medium (DUO82040, Duolink) and imaged using a Nikon Eclipse TE2000-U inverted microscope equipped with an EMCCD camera. The number of discrete fluorescent puncta within a randomly chosen 20μm axon segment in each image was counted using Volocity.

49Super-resolution 3D SIM images were acquired using a Deltavision OMX 3D SIM System V3 from Applied Precision (a GE Healthcare company) equipped with 3 EMCCD Cascade cameras from Photometrics, 488nm and 592.5 nm diode laser illumination, an Olympus PlanSApo 100 × 1.40 NA oil objective, and standard excitation and emission filter sets. Imaging of each channel was performed sequentially using three angles and five phase shifts of the illumination pattern as described. The refractive index of the immersion oil (Cargille) was adjusted to 1.513 to minimize spherical aberrations. Sections were acquired at 0.1 μm z steps. Raw OMX data were reconstructed and channel registered in the SoftWoRx software (Applied Precision, a GE Healthcare company). Reconstructions were carried out using channel-specific Optical Transfer Functions (OTFs) and channel-specific K0 angles. OTFs were generated within the SoftWoRx software by imaging 100 nm beads (Life Technologies) using appropriate immersion oils to match the data. Channel registration was carried out using the Image Registration parameters generated within the SoftWoRx software and checked for accuracy by imaging Tetraspeck beads (Life Technologies). Channel registration was accurate to one pixel. Further data analysis was performed using Fiji.

50Anaesthetized embryos of either sex were fixed at stage 39 or 45 in 4% formaldehyde in PBS at 4°C overnight and washed 3 times with filtered PBS. RGC axons of one eye per embryo were labeled by intraocular injection of fluorescent lipophilic dye, DiI (ThermoFisher Scientific). Embryos were left in PBS for 48 hours allowing the diffusion of the dye into the optic tract. The contralateral (or non-dye injected side) brain hemisphere was later dissected, mounted in PBS and visualized using confocal under 559nm laser line. Brains with less than 30 visible axons were categorized as defective axonal projection phenotypes. For single-blastomere injection experiments, axonal projections were analyzed by using Volocity automated detection of the fluorescence signal, creating a 3D mask matching the signal and providing the corresponding volume.

51All experiments were performed in at least three independent biological replicates. The n number for each experiment, details of statistical analysis and software are described in the figure legends or main text. Statistical analyses used in this study include Wilcoxon Ranksum Test, Mann-Whitney test and Fisher’s exact Test. Statistical significance is defined as, n.s., not significant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Statistical analysis was performed using Prism (GraphPad) or MATLAB (MathWorks).